Research map / Federated learning and privacy / Communication-efficient FL
Federated Deep AUC Maximization for Hetergeneous Data with a Constant Communication Complexity
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Add primal and dual control variates to local updates | CODA+ framework (CODASCA) · ImageNet-IH (10%), K=16 | maximum communication window I without major performance loss: 32 -> 128 | Helped |
| Add primal and dual control variates to local updates | CODA+ framework (CODASCA) · ImageNet-IH (10%), K=8 | maximum communication window I without major performance loss: 128 -> 512 | Helped |
| Add primal and dual control variates to local updates | CODA+ framework (CODASCA) · ChestXray-IH, C0 | AUC: 0.8361 -> 0.8427 | Helped |
| Add primal and dual control variates to local updates | CODA+ framework (CODASCA) · ChestXray-IH, C2 | AUC: 0.7356 -> 0.7401 | Helped |
| Add primal and dual control variates to local updates | CODA+ framework (CODASCA) · ChestXray-IH, C3 | AUC: 0.7449 -> 0.7468 | Helped |
| Add primal and dual control variates to local updates | CODA+ framework (CODASCA) · ChestXray-IH, C1 | AUC: 0.8464 -> 0.8457 | Hurt |
| Add primal and dual control variates to local updates | CODA+ framework (CODASCA) · ChestXray-IH, C4 | AUC: 0.8680 -> 0.8680 | No clear effect |
| Train with additional organizations' data rather than CheXpert data alone | CODASCA with DenseNet121 · CheXpert validation, AVG | AUC: 0.9353 -> 0.9376 | Helped |
| Train with additional organizations' data rather than CheXpert data alone | CODASCA with DenseNet161 · CheXpert validation, AVG | AUC: 0.9316 -> 0.9370 | Helped |
| Remove the extra post-stage large-batch computation of the dual variable | CODA, yielding CODA+ · Federated deep AUC maximization | sample complexity | Helped |
| Use stochastic controlled averaging to correct client drift | CODASCA · Heterogeneous federated deep AUC maximization | communication complexity: ˜O(K/μ + 1/(με¹ᐟ²) + 1/(μ³ᐟ²ε¹ᐟ²)) -> ˜O(1/μ) | Helped |
| Increase the number of data sources from one to five | CODASCA with DenseNet121 · CheXpert validation, C0 | AUC: 0.9007 -> 0.9066 | Helped |
Closest papers by meaning
- A Federated Stochastic Multi-level Compositional Minimax Algorithm for Deep AUC Maximization (ICML 2024 · similarity 0.89)
- Federated Compositional Deep AUC Maximization (NeurIPS 2023 · similarity 0.87)
- Solving a Class of Non-Convex Minimax Optimization in Federated Learning (NeurIPS 2023 · similarity 0.85)
- FedVCK: Non-IID Robust and Communication-Efficient Federated Learning via Valuable Condensed Knowledge for Medical Image Analysis (AAAI 2025 · similarity 0.85)
- FedRTS: Federated Robust Pruning via Combinatorial Thompson Sampling (NeurIPS 2025 · similarity 0.83)
- MH-pFLID: Model Heterogeneous personalized Federated Learning via Injection and Distillation for Medical Data Analysis (ICML 2024 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.